GIScience & Remote Sensing

Remote Sensing-Assisted Foundation Models for Geosciences

Foundation models are transforming the landscape of remote sensing and geoscientific research by enabling large-scale, transferable, and multimodal understanding of the Earth system. Pretrained on vast heterogeneous datasets, these models unify information from diverse modalities such as optical, synthetic aperture radar, hyperspectral, digital elevation model, and atmospheric observations to support a wide range of spatial and temporal analyses. Their growing integration with geospatial information science allows for new capabilities in cross-modal reasoning, open-vocabulary mapping, and generative environmental modeling, paving the way toward more adaptive and high-performing geospatial intelligence. At the same time, these advances call for systematic benchmarking, efficient fine-tuning, and trustworthy reasoning mechanisms that ensure effectiveness, robustness and transparency.

Remote sensing foundation models marks a pivotal shift in how geoscientists analyze and interpret Earth observation data. Traditional task-specific models often struggle with data scarcity, regional bias, and limited generalization across sensors or scales. Foundation models, by contrast, leverage large-scale pretraining and multimodal fusion to capture richer spatial–temporal representations that transfer effectively across diverse environments. This capability is crucial for addressing pressing challenges in climate monitoring, environmental change detection, disaster management, and sustainable development, where decisions increasingly rely on fast, reliable, and explainable geospatial intelligence. Moreover, the convergence of these models with large language models, agentic reasoning, and knowledge-grounded frameworks introduces opportunities for human–AI collaboration, automated interpretation, and transparent decision support. Advancing research in this area is therefore essential to build high-performing, robust, and trustworthy systems that can transform remote sensing data into reliable geoscientific discovery for various geoscience applications.

This Article Collection aims to highlight cutting-edge progress in remote sensing–assisted foundation models and their applications in geoscience, fostering collaboration across Earth observation, GIScience, and artificial intelligence communities. We invite original research articles and reviews that align with the aims and scope of GIScience & Remote Sensing, emphasizing the intersection of foundation models, geospatial intelligence, and geoscientific applications.

Submissions are encouraged in three complementary areas.

The first focuses on benchmarking remote sensing foundation models for geoscience datasets and tasks, including multimodal LLMs, diffusion models, and representation-based vision foundation models for both high-level analytical and low-level perceptual problems.

The second explores efficient fine-tuning strategies for adapting foundation models, such as Skysense, DOFA, Prithvi, and AlphaEarth embeddings, to diverse GIS and Earth system applications.

The third highlights agentic geospatial and geoscientific reasoning, integrating memory, retrieval-augmented generation, and knowledge graphs for autonomous spatial analysis and decision support. Contributions emphasizing reproducibility, open data and benchmarks, interpretability, and real-world deployment are particularly welcome to advance transparent, interdisciplinary progress in next-generation GIScience and remote sensing research.

Article Collection Guest Advisors

Dr. Zhitong Xiong is currently a Post-Doctoral Researcher with the Chair of Data Science in Earth Observation, Technical University of Munich (TUM), Munich, Germany. His research interests include machine learning, remote sensing, EO foundation models, and multimodal fusion.

Professor Muhammad Shahzad is a Lecturer at the University of Reading, UK, specializing in Radar Remote Sensing with a focus on 3D object reconstruction using SAR imagery. His research spans 3D computer vision and image analysis, particularly in developing deep learning methods for processing multimodal remote sensing data, including unstructured and structured 3D point clouds, optical RGB-D data, and very high-resolution radar images.

Dr. Yuchi Ma is a postdoctoral scholar in Earth System Science at Stanford University. His research integrates remote sensing, agricultural domain knowledge, and AI to enhance understanding of agroecosystems and inform agricultural policymaking. His work is dedicated to enhancing global food security and promoting agricultural resilience through the development and integration of multidisciplinary technologies.

­­All manuscripts submitted to this Article Collection will undergo a full peer-review; the Guest Advisors for this Collection will not be handling the manuscripts.

Please review the journal scope and author submission instructions prior to submitting a manuscript.

The deadline for submitting manuscripts is 31 July 2026

Please contact Commissioning Editor Alex Johnson at [email protected] with any queries and discount codes regarding this Article Collection.

Please be sure to select the appropriate Article Collection from the drop-down menu in the submission system.

Article Collection Key Terms:

1. Remote sensing foundation models
2. Geoscientific reasoning with foundation models
3. Benchmarking foundation models for geoscience
4. Efficient Fine-tuning of foundation models
5. Foundation models for geoscientific applications

The Guest Advisors have declared no conflict of interests in preparing this Article Collection